JPKCom DB

watch — Run a Command Repeatedly and Watch Its Output

Practical guide to watch — repeat commands at fixed intervals and follow changing output live, with custom intervals, diff highlighting and exit conditions.

watch runs a command over and over at a fixed interval and shows you only its latest output – perfect for keeping an eye on changing data such as disk usage, process lists, network connections or the status of a deployment. By default it repeats every two seconds; a handful of flags let you change the interval, highlight what changed or stop the loop automatically. This guide walks you through the options and patterns you actually reach for day to day.

Basic Usage

watch <command> — Run a command every 2 seconds (default interval).

watch date

watch -n <seconds> <command> — Run at a custom interval.

watch -n 5 df -h

watch -n 0.5 <command> — Run every half second (sub-second intervals).

watch -n 0.5 'cat /proc/loadavg'

watch '<command1> | <command2>' — Watch a piped command (use quotes).

watch 'ps aux | grep nginx | wc -l'

Display Options

watch -d <command> — Highlight differences between updates.

watch -d free -m

watch -d=permanent <command> — Highlight all changes since the first run (cumulative).

watch -d=permanent 'ls -la /tmp'

watch -t <command> — Hide the header (no title, interval, or command shown).

watch -t date

watch --color <command> — Preserve ANSI color codes in output.

watch --color 'ls --color=always -la'

Exit Conditions

watch -g <command> — Exit when the output changes.

watch -g 'ls /tmp/upload/'

watch -e <command> — Exit on error (freeze output and wait for keypress).

watch -e 'curl -sf http://localhost:8080/health'

watch -b <command> — Beep when the command returns a non-zero exit code.

watch -b 'ping -c 1 -W 1 server.example.com'

Common Patterns

watch -n 1 'kubectl get pods' — Monitor Kubernetes pod status.

watch -n 1 'kubectl get pods -n production'

watch -d 'df -h' — Monitor disk usage with change highlighting.

watch -d 'df -h'

watch -n 2 'docker ps' — Monitor running Docker containers.

watch -n 2 'docker ps --format "table {{.Names}}	{{.Status}}"'

watch -n 5 'ss -tlnp' — Monitor listening ports.

watch -n 5 'ss -tlnp'

watch -n 1 'cat /proc/net/dev' — Monitor network interface traffic in real-time.

watch -n 1 'cat /proc/net/dev'

watch -g 'checksum.sh' — Wait until a file changes (exit on change).

watch -g 'md5sum deploy.tar.gz'

watch -n 10 'curl -so /dev/null -w "%{http_code}" <url>' — Monitor HTTP status code of an endpoint.

watch -n 10 'curl -so /dev/null -w "%{http_code}" http://localhost:8080/health'

Conclusion

watch turns any status command into a live dashboard without you having to write a while loop – for everyday use you rarely need more than -n for the interval, -d to highlight changes and -g to bail out automatically when something changes. Keep two things in mind. As soon as pipes, variables or glob patterns are involved, wrap the command in quotes (watch '…') or run it via watch bash -c '…', otherwise your current shell expands it once and watch just repeats the frozen result. And because watch runs the command again and again, only ever feed it read-only status and query commands – never anything that changes or deletes data.

Further Reading

  • top – interactive real-time view of processes and system load
  • ps – snapshot of the currently running processes
  • journalctl – query and live-follow the systemd journal logs